Sparse Optimization Techniques for Differential Equation Discovery
When
1:45 to 2:30 p.m., Nov. 8, 2023
Where
Speaker: Teddy Meissner, Program in Applied Mathematics, University of Arizona
Abstract: Differential equations are essential for modeling in many scientific areas, but often we're working with partial models or sometimes none at all due to gaps in our understanding. I will discuss a method to discover models from noisy data by simultaneously employing an $L_0$ approximation to identify functional subsets and the Akaike Information Criterion (AIC) to accept or reject these steps. I will also discuss sparse automatic differentiation techniques, which are crucial for large-scale second order optimization.